A Dynamic Cluster Model Based on Projection Pursuit with Its Application to Climate Zoning

Climate zoning analysis is a typical multifactor problem.The difficulty frequently encountered in climate zoning analysis is that there are so many factors and the complex interrelationship among them cannot be analyzed according to only one factor, all the effect factors associated with climate zoning must be taken into consideration. Aiming at the problem mentioned above, a dynamic cluster model based on projection pursuit principle (DCPP), in which dynamic cluster is combined with projection pursuit principle, is developed, and it is used in climate zoning successfully for the first time. Firstly, multifactor cluster problem can be converted into singlefactor (projected characteristic value) cluster problem according to linear projection principle. Secondly, a new projection index based on dynamic cluster rule is constructed in the dynamic cluster model based on projection pursuit principle, which is the clustering basis for the projected characteristic value. Thirdly, genetic algorithm (GA) is applied to optimize the dynamic cluster model based on projection pursuit principle, and the steps of genetic algorithm are introduced in detail. Finally, a case study on climate zoning is used to test the effect of the dynamic cluster model based on projection pursuit principle.The results show that the dynamic cluster model based on projection pursuit principle for climate zoning is reasonable and effective. On the other hand, based on the dynamic cluster model based on projection pursuit principle, the cluster results can be obtained directly according to the characteristic of data set. Since there is no parameter calibration in the dynamic cluster model based on projection pursuit principle, the results are more objective and less subjective.The dynamic cluster model based on projection pursuit principle is a new method and powerful tool in climate zoning. A new approach to the problem of complicated multifactor cluster analysis is provided by the dynamic cluster model based on projection pursuit principle.